Linear regression analyses with dichotomous MRI measures (separate analyses for white matter hyperintensities, lacunes, and microbleeds) and cerebrospinal fluid biomarker levels of β-amyloid 42 (Aβ42) (A), tau (B), and tau phosphorylated at threonine 181 (P-tau181) (C) (dependent variables) were used, adjusted for age, sex, mediotemporal lobe atrophy, and diagnosis (as categories). Interaction terms for diagnosis MRI measure were included to estimate effect sizes per diagnosis group. Bars indicate standardized betas; error bars, 95% confidence intervals of the standardized betas.aP < .05.bP < .005.
Kester MI, Goos JDC, Teunissen CE, Benedictus MR, Bouwman FH, Wattjes MP, Barkhof F, Scheltens P, van der Flier WM. Associations Between Cerebral Small-Vessel Disease and Alzheimer Disease Pathology as Measured by Cerebrospinal Fluid Biomarkers. JAMA Neurol. 2014;71(7):855-862. doi:10.1001/jamaneurol.2014.754
Copyright 2014 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
It remains unclear if and how associations between cerebral small-vessel disease and Alzheimer disease (AD) pathology lead to cognitive decline and dementia.
To determine associations between small-vessel disease and AD pathology.
Design, Setting, and Participants
Cross-sectional study from January 2002 to December 2012 using the memory clinic–based Amsterdam Dementia Cohort. The study included 914 consecutive patients with available cerebrospinal fluid (CSF) and magnetic resonance imaging; 547 were patients diagnosed as having AD (54% female, mean [SD], 67 ; Mini-Mental State Examination score, mean [SD], 21 ), 30 were patients diagnosed as having vascular dementia (37% female, mean [SD], 76 ; Mini-Mental State Examination score, mean [SD], 24 ), and 337 were control participants with subjective memory complaints (42% female, mean [SD], 59 ; Mini-Mental State Examination score, mean [SD], 28 ). Linear regressions were performed with CSF biomarkers (log transformed) as dependent variables and magnetic resonance imaging measures (dichotomized) as independent, adjusted for sex, age, mediotemporal lobe atrophy, and diagnosis. An interaction term for diagnosis by magnetic resonance imaging measures was used for estimates per diagnostic group.
Main Outcomes and Measures
We examined the associations of magnetic resonance imaging white matter hyperintensities (WMH), lacunes, microbleeds with CSF β-amyloid 42 (Aβ42), total tau, and tau phosphorylated at threonine 181 (P-tau181) as well as for a subset of apolipoprotein E (APOE) ε4 carriers and noncarriers.
Microbleed presence was associated with lower CSF Aβ42 in AD and vascular dementia (standardized beta = −0.09, P = .003; standardized beta = −0.30, P = .01), and higher CSF tau in controls (standardized beta = 0.10, P = .03). There were no effects for P-tau181. The presence of WMH was associated with lower Aβ42 in control participants and patients with vascular dementia (standardized beta = −0.18, P = .002; standardized beta = −0.32, P = .02) but not in patients with AD. There were no effects for tau or P-tau181. The presence of lacunes was associated with higher Aβ42 in vascular dementia (standardized beta = 0.17, P = .07) and lower tau in AD (standardized beta = −0.07, P = .05) but there were no effects for Aβ42 or P-tau181. Stratification for apolipoprotein E genotype revealed that these effects were mostly attributable to ε4 carriers.
Conclusions and Relevance
Deposition of amyloid appears aggravated in patients with cerebral small-vessel disease, especially in apolipoprotein E ε4 carriers, providing evidence for pathophysiological synergy between these 3 biological factors.
Quiz Ref IDAlzheimer disease (AD) is thought to be caused by amyloid aggregation and the formation of tau tangles.1 By contrast, in vascular dementia (VaD), infarcts or profuse white matter disease are considered the cause of cognitive decline.2 However, postmortem studies show a high prevalence of mixed pathology, especially in elderly populations.3 Many studies show cerebral small-vessel disease (SVD) and vascular risk factors to increase the risk of developing AD,4,5 while in both VaD and SVD, signs of AD pathology have been reported.6- 8
It remains unclear how the interaction between SVD and AD pathology leads to dementia. Small-vessel disease could be an independent process leading to dementia in combination with coexisting yet unrelated AD pathology, but ischemic SVD might also contribute directly to AD pathology, eg, by accelerating the rate of amyloid deposition owing to ischemic changes or by inducing ischemia owing to amyloid deposition in the vessels (cerebral amyloid angiopathy [CAA]).9- 12Quiz Ref IDThe apolipoprotein E (APOE) ε4 genotype is a well-known risk factor not only for AD but also for cardiovascular disease and CAA, and it could be a modifying factor in the association of SVD and AD.13- 17
Cerebrospinal fluid (CSF) β-amyloid 42 (Aβ42), total tau (tau), and tau phosphorylated at threonine 181 (P-tau181) are considered to reflect AD pathophysiology.18 White matter hyperintensities (WMH) and lacunes are magnetic resonance imaging (MRI) measures for ischemic SVD.2 Microbleeds (MBs) are considered a marker of underlying CAA and have also been related to hypertensive vasculopathy.9,19 By examining relationships of MRI-based MBs, WMH, and lacunes with levels of CSF Aβ42, tau, and P-tau181 in patients with AD, patients with VaD, and control participants, we explored the relationship between SVD and AD pathology. In addition, we examined the modifying effects of the APOE genotype on these relations.
We included 914 consecutive patients in our study; 547 were patients diagnosed as having AD, 30 were patients diagnosed as having VaD, and 337 were control participants with subjective memory complaints from the memory clinic–based Amsterdam Dementia Cohort with available data on their CSF and MRI. All patients underwent a standard dementia screening including physical and neurological examination as well as laboratory tests, electroencephalogram, and brain MRI. Cognitive screening included Mini-Mental State Examination (MMSE) and usually involved comprehensive neuropsychological testing. The diagnosis of probable AD was made according to the National Institute of Neurological and Communicative Disorders and Stroke–Alzheimer’s Disease and Related Disorders Association criteria20 and all patients fulfilled core clinical criteria according to National Institute on Aging–Alzheimer’s Association.21 Diagnosis of VaD was made according to the National Institute of Neurological Disorders and Stroke–Association Internationale pour la Recherche et l'Enseignement en Neurosciences criteria.2 When results of all examinations were normal, patients were considered to have subjective memory complaints. Patients with mild cognitive impairment22 or with a psychiatric disorder were not included. Diagnoses were made by consensus of a multidisciplinary team, without knowledge of CSF results or APOE genotype. Patients were asked about education, current use of medication, alcohol use, smoking history, and medical history. Diabetes mellitus was defined as use of glucose-lowering agents or known history of diabetes, hypertension as use of antihypertensive agents or known history of hypertension, hypercholesterolemia as known history of hypercholesterolemia, and myocardial infarction as known history of myocardial infarction. The local institutional review board at the VU University Medical Center approved the study and all patients provided written informed consent.
Scans were obtained on 4 different MRI scanners (in order of frequency): Signa 3.0-T (n = 557) (GE), Magnetom Impact 1.0-T (n = 243) (Siemens), Sonata 1.5-T (n = 77) (Siemens), Signa 1.5-T (n = 20) (GE), and Avanto 1.5-T (n = 16) (Siemens). The MRI rater was a neuroradiologist with special dementia neuroimaging expertise and was blind to clinical data. Microbleeds were defined as rounded hypointense homogeneous foci up to 10 mm in the brain parenchyma on T2*-weighted images. On the fluid-attenuated inversion recovery (FLAIR) sequence, WMH were assessed using the Fazekas scale (none, punctuate, early confluent, confluent; score 0-3).23 Lacunes were defined as deep lesions from 3 to 15 mm with CSF-like signal on FLAIR, T1-weighted, and T2-weighted images. Medial temporal lobe atrophy (MTA) was rated (0-4) using oblique coronal reconstructions of T1-weighted gradient-echo volume sequences perpendicular to the long axis of the hippocampus.24 Global cortical atrophy was assessed on the axial FLAIR sequence (0-3).25 On both scales, maximal atrophy is represented by the highest score. Magnetic resonance imaging readings were dichotomized as the following: MBs, 0 vs 1 or more; WMH, 0 or 1 vs 2 or 3; lacunes, 0 vs 1 or more; and MTA, (average score left and right side) 0 or 1 vs 2 or more.
Cerebrospinal fluid was obtained by lumbar puncture using a 25-gauge needle and was collected in 10-mL polypropylene tubes. Within 2 hours, CSF samples were centrifuged at 1800g for 10 minutes at 4°C. A small amount of CSF was used for routine analysis including total cells, total protein levels, and glucose levels. Cerebrospinal fluid supernatant was transferred to 0.5- or 1-mL polypropylene tubes and stored at −20°C for enzyme-linked immunosorbent assay analysis within 1 month. Biomarkers of CSF were measured with Innotest sandwich enzyme-linked immunosorbent assay as previously described.26 As the manufacturer does not supply controls, the performance of the assays was monitored with pools of surplus CSF specimens. The mean (SD) intra-assay coefficient of variation was 2.0% (0.5%) for Aβ42, 3.2% (1.3%) for tau, and 2.9% (0.8%) for P-tau181 as calculated from averaging the CV of duplicates from 5 runs randomly selected over 2 years. The mean (SD) interassay coefficient of variation was 10.9% (1.8%) for Aβ42, 9.9% (2.1%) for tau, and 9.1% (1.8%) for P-tau181, as analyzed in a high and low pool from 13 consecutive pool preparations used in total in 189 to 231 runs. The team involved in the CSF analysis was not aware of the clinical diagnoses.
For APOE genotyping, DNA was isolated from 10 mL of EDTA blood by the QIAamp DNA blood isolation kit (Qiagen). The genotype was determined with the Light Cycler APOE mutation detection kit (Roche Diagnostics GmbH). Patients were classified as APOE ε4 noncarriers or carriers. Data on APOE were available for 861 patients (94%) (327 control participants, 507 patients with AD, and 27 patients with VaD).
Biomarker levels of CSF were log transformed because they were not normally distributed. Differences between diagnostic categories for baseline characteristics were assessed using analysis of variance with post hoc Bonferroni corrections or Fisher exact test when applicable. Linear regression analyses were used to investigate the association between dichotomous MRI measures (independent variables) and CSF biomarker levels of Aβ42, tau, and P-tau181 (dependent variables), adjusted for age, sex, MTA, diagnosis (as categories), and interaction terms for diagnosis by MRI measure to estimate effect sizes per diagnostic group. First, separate analyses were done for MBs, WMH, and lacunes. Second, we combined the different MRI measures in 1 multivariate model to analyze independent effects on the CSF biomarker levels.
Finally, we investigated the effect of APOE ε4 on the associations between the MRI measures and CSF biomarkers by adding APOE ε4 genotype and interaction terms for (1) diagnosis by MRI measures, (2) APOE ε4 genotype by MRI measure, and (3) APOE ε4 genotype by MRI measure by diagnosis. If there was interaction (APOE ε4 genotype by MRI measure, P <.10), we reported the associations between MRI measures and CSF biomarkers separately for APOE ε4 noncarriers and carriers. We used the third-grade interaction term (APOE ε4 genotype by MRI measure by diagnosis) to estimate effect sizes per diagnosis group. If there was no significant interaction, the interaction term was removed from the model and we reported with APOE ε4 genotype as covariate. In all analyses, the standardized betas are given with a P value. We used SPSS version 21 statistical software (IBM) for analyses.
Baseline characteristics are shown in Table 1. Cerebrospinal fluid Aβ42 levels were lower in patients with AD than in control participants and patients with VaD; in addition, CSF Aβ42 levels were lower in patients with VaD than in control participants. Quiz Ref IDLevels of CSF tau and P-tau181 were higher in patients with AD than in both control participants and patients with VaD. Carrying APOE ε4 was more common in patients with AD than in control participants. Microbleeds, lacunes, and WMH were more common in patients with VaD than in both control participants and patients with AD; in addition, MBs and WMH were more common in patients with AD than in control participants.
Linear regression analyses adjusted for sex, age, and MTA showed that vascular MRI measures were mainly associated with CSF Aβ42 levels (Figure 1). The presence of MBs was associated with lower Aβ42 in patients with AD and VaD, with a trend in control participants (Table 2). Control participants with MBs had higher levels of tau. There were no effects for P-tau181. The presence of WMH was associated with lower Aβ42 in control participants and patients with VaD, with only a trend in patients with AD, but there were no effects for tau or P-tau181. For presence of lacunes, there were only trends: the presence of lacunes was associated with higher Aβ42 in patients with VaD (P = .07) and lower tau in patients with AD (P = .05).
When we analyzed all 3 MRI measures in 1 combined model, most associations remained the same (Table 2). Associations between MBs and Aβ42 in patients with AD as well as between MBs and higher tau in control participants just lost significance, but effect sizes remained in the same order of magnitude. Contrary to results of separate models, lacunes were the strongest predictor for higher (ie, less reduced) CSF Aβ42 in patients with VaD, whereas WMH no longer predicted levels of Aβ42 in patients with VaD. While it was not associated in the separate model, WMH were associated with lower levels of CSF P-tau181 in patients with AD in the combined model.
We included APOE in the model in an additional set of analyses. When interaction was significant, associations are reported separately for APOE ε4 carriers and noncarriers, and if there was no interaction, we show results adjusted for APOE genotype (Table 3). In 4 models there was an interaction between the MRI measure and APOE genotype (P < .10). In control participants and patients with AD the association between MBs and lower Aβ42 was seen only in APOE ε4 carriers. In patients with AD there appeared to be an association between MBs and both lower tau and P-tau181 in APOE ε4 noncarriers but not in carriers. In control participants there was an association between WMH and lower CSF Aβ42 in APOE ε4 carriers, but not in noncarriers. In the other models there was no significant interaction with APOE genotype, and adjustment for APOE genotype did not change results in these analyses.
Quiz Ref IDIn our study, the presence of both MBs and WMH was associated with lower CSF levels of Aβ42, indicating a direct relationship between SVD and AD pathology. Amyloid pathology appears aggravated in patients with vascular damage, which supports pathophysiological synergy. In control participants, tau levels were elevated in the presence of MBs, which could indicate that this increase of tau is a result of neuronal cell death in patients not diagnosed as having AD. Effects on CSF Aβ42 levels were largest in ε4 carriers, indicating an inducing role of the APOE ε4 genotype in the relation between AD and vascular pathology.
Previous studies have shown close relations among MBs, CAA, and amyloid pathology. Hypotheses for this close relation include that the abundance of parenchymal amyloid, as seen in patients with AD, leads to more vascular amyloid, CAA, and MBs9,11 and that vessel brittleness leads to reduced clearance of amyloid, also resulting in CAA, MBs, and an increase of amyloid pathology.7,27,28 This is in line with the association between lower Aβ42 and MBs in both patients with AD and patients with VaD in our study. Furthermore, it has been suggested that amyloid accumulation in small arteries and capillaries leads to vessel wall changes, which in turn leads to lumen obstruction and ischemia.10- 12,29,30 This is in line with the associations we found between lower Aβ42 levels and WMH in both control participants and patients with VaD. It is conceivable that CAA, which is commonly seen in patients with AD and is associated with parenchymal amyloid, may lead to ischemic vascular events in the brain like WMH, microinfarcts, and MBs.10- 12 Conversely, ischemic changes could accelerate the rate of amyloid deposition, and vessel wall stiffness may impair perivascular drainage of cerebral amyloid, both leading to more deposition.9
Lacunes seem to reflect a different type of pathology. We found a positive association between Aβ42 and lacunes in patients with VaD, indicating that lacunes are associated with lower amounts of amyloid in the brain. Lacunes and AD pathology seem to lead independently to cognitive decline, while a combination of enough additive damage leads to clinical (vascular) dementia.3 The difference between the associations of amyloid and both MBs and WMH on one hand and lacunes on the other hand could be related to anatomy. Decreased amyloid clearance and vessel wall brittleness seem to be problems of the smallest arteries and capillaries of the cortex, mainly due to CAA, while lacunes are caused by obstruction of the larger arterioles, mainly in deeper regions, rather than associated with hypertensive vasculopathy.31
The APOE ε4–noncarrying patients with AD with MBs had lower levels of both tau and P-tau181 compared with those without MBs. Tau tangle pathology seems relatively unrelated to vascular pathology and could independently lead to cognitive decline. We hypothesize that owing to a relatively higher degree of vascular damage in this group of patients, a relatively lower degree of tau tangle pathology is sufficient for a diagnosis of (AD) dementia. For amyloid pathology, effects were the other way around. Our study showed that APOE ε4 carriers with MBs or WMH showed lower levels of CSF Aβ42, which is in line with previous studies.5,13- 17,32,33 The APOE ε4 genotype is a risk factor for both AD and CAA.13- 17 Moreover, it is also associated with increased inflammatory response,34 less effective degradation of amyloid,35 and reduced transmembrane amyloid transport.13 In overly expressed amyloid, there will be more CAA in ε4 carriers owing to impaired mechanisms of amyloid clearance in this genotype.33 In turn, the accumulated CAA could lead to more MBs and WMH owing to increased inflammation. Our results support the hypothesis that SVD and AD pathology are directly associated. The APOE ε4 genotype seems to upregulate this association (Figure 2).
Previous studies have not consistently shown associations of WMH with CSF biomarkers or Pittsburgh Compound B amyloid load, possibly owing to lack of power as most of these studies had a relatively small sample size.27,29,36- 38 At the center where this study took place, lumbar puncture and MRI are both performed on a routine basis in most patients presenting with memory problems, resulting in a large study sample. We did not account for WMH pathology location, although there are studies suggesting that either frontally and/or parieto-occipitally located WMH are associated more with neurofibrillary pathology.27,39 It has been suggested that deep WMH are less associated with amyloid pathology, as deep regions can also diffuse amyloid directly into the CSF (via the ventricles).27 Others have hypothesized that frontal WMH are more heterogeneous and do not always reflect ischemia but could also reflect cortical neurodegeneration.39 Further (postmortem) studies are needed to unravel this. Another possible limitation of our study is that we had a relatively small number of patients with VaD, making the estimated effects less reliable (ie, larger confidence intervals) for this group. However, pure VaD is a rare disorder and large data sets are difficult to attain.
Quiz Ref IDOur study supports the hypothesis that the pathways of SVD and AD pathology are interconnected. Small-vessel disease could provoke amyloid pathology while AD-associated cerebral amyloid pathology may lead to auxiliary vascular damage. Treatment trials for vascular risk factors should take amyloid reduction into account, especially in APOE ε4 carriers. Evaluation of effects of amyloid-reducing therapies like immune-directed therapies should be stratified for the presence of MBs and WMH, and not only as safety measure,40 as effects of amyloid reduction could be affected by the presence of cerebral SVD pathology.
Corresponding Author: Maartje I. Kester, MD, PhD, Alzheimer Center, Department of Neurology, VU University Medical Center, PO Box 7057, 1007 MB Amsterdam, the Netherlands (firstname.lastname@example.org).
Accepted for Publication: March 19, 2014.
Published Online: May 12, 2014. doi:10.1001/jamaneurol.2014.754.
Author Contributions: Drs Kester and van der Flier had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Kester, Wattjes, Scheltens, van der Flier.
Acquisition, analysis, or interpretation of data: Kester, Goos, Teunissen, Benedictus, Bouwman, Wattjes, Barkhof, Scheltens.
Drafting of the manuscript: Kester, Wattjes.
Critical revision of the manuscript for important intellectual content: Goos, Teunissen, Benedictus, Bouwman, Wattjes, Barkhof, van der Flier.
Statistical analysis: Kester, van der Flier.
Obtained funding: Kester, Scheltens.
Administrative, technical, or material support: Benedictus, Scheltens.
Study supervision: Teunissen, Wattjes, Barkhof, Scheltens, van der Flier.
Conflict of Interest Disclosures: Dr Teunissen serves as a member of the scientific advisory board of Innogenetics SA and Roche; received a speaker honorarium on a Tea-sponsored symposium; and received grants from the European Commission and the Alzheimer’s Drug Discovery Foundation. Dr Wattjes received consulting fees from Biogen Idec and Roche and he serves as an editorial board member of European Radiology. Dr Barkhof serves and has served on the advisory boards of Bayer-Schering Pharma, sanofi-aventis, Biogen Idec, Teva, Merck-Serono, Novartis, Jansen, and Roche; has received funding from the Dutch MS Society; and has been a speaker at symposia organized by the Serono Symposia Foundation. Dr Scheltens serves and has served on the advisory boards of Genentech, Novartis, Roche, Danone, Nutricia, Lilly, and Lundbeck. He has been a speaker at symposia organized by Lundbeck, Merz, Danone, Novartis, Roche, GE, and Genentech and receives no financial compensation for his activities. No other disclosures were reported.
Funding/Support: Dr Kester was supported by research fellowship WE 15-2012-03 from Alzheimer Nederland. Research performed at the Alzheimer Center, VU University Medical Center is part of the neurodegeneration research program of Neuroscience Campus Amsterdam. The Alzheimer Center, VU University Medical Center is supported by Alzheimer Nederland and the Stichting VU University Medical Center funds. The clinical database structure was developed with funding from Stichting Dioraphte.
Role of the Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Correction: This article was corrected on May 22, 2014, to fix a typographical error in the Abstract.